# a2a (blogs) — RSS Amplifier Recent posts from the 2 feeds in the RSS Amplifier directory that cover a2a. Page: Feed: --- ## [What your evals never read](https://thepragmaticquant.com/what-your-evals-never-read/) _2026-08-06 · thepragmaticquant_ Your eval suite scores the function. It never reads the sentence the model reads before deciding to call that function, and that sentence is the one that can lie, be tampered with after you ship, or be obeyed differently on the very next run. First in a series on what agent evals actually have to measure. ## [Nobody published the denominator](https://thepragmaticquant.com/nobody-published-the-denominator/) _2026-07-30 · thepragmaticquant_ AI-for-security benchmarks that score models on compiled code can report a number nobody can recompute. Of 37 papers reporting an obfuscation-strength number, exactly 1 conditions it on whether the transform actually fired — and that one is a disassembly paper, which scores nothing on re-executability. Our own do-nothing control passes the metric. ## [The compiler did it, and my obfuscation took the credit](https://thepragmaticquant.com/the-compiler-did-it/) _2026-07-30 · thepragmaticquant_ Two arms of the same 149 functions, neither arm obfuscated — stock LLM4Decompile v1.5, gcc-lenient scorer, native 4096-token window: GCC-built binaries recover at 27.5%, clang-built at 15.4%, paired McNemar p = 0.0029. The confound is the toolchain, and the law generalizes well past binary analysis. The three-term decomposition built to separate those effects then fails to add up, and the… ## [Bigger didn't break it. Knowing did.](https://thepragmaticquant.com/bigger-didnt-break-it/) _2026-07-30 · thepragmaticquant_ Obfuscation dropped this decompiler's recovery from 18.8% of clean functions (28/149) to 0.0-0.9% of the functions each transform fired on (0/97, 0/63, 1/106) — gcc-lenient scorer, native 4096-token window, paired and placebo-controlled. Tripling its parameters bought no monotonic gain in the largest obfuscated cell: three points at n = 151, intervals overlapping. Fine-tuning it on the same… ## [mcp 1.x to 2.0: reference notes](https://thepragmaticquant.com/mcp-1-to-2-reference-notes/) _2026-07-29 · thepragmaticquant_ Every breaking change in the Python MCP SDK's 2.0, what it looks like when it bites you, and what to do about it. Then why the protocol changed the way it did. Measured against installed 1.26.0 and 2.0.0 wheels. ## [Ten thousand series, one pass](https://thepragmaticquant.com/ten-thousand-series-one-pass/) _2026-07-15 · thepragmaticquant_ Ten thousand series, a thousand replicates each, and the naive tensor wants 16 GB before the first statistic lands. This is the real race: three implementations, two axes, and each speedup credited to the opponent it actually outran. ## [Count the bytes, not the FLOPs](https://thepragmaticquant.com/why-we-stopped-materializing-arrays/) _2026-07-11 · thepragmaticquant_ Why tsbootstrap stopped materializing arrays and carrying state. A routine bootstrap job can materialize a 160 MB tensor whose only purpose is to be averaged and thrown away. How tsbootstrap's hot path stopped building it, including the one path where the incumbent still wins. ## [When your errors aren’t equal](https://thepragmaticquant.com/when-your-errors-arent-equal/) _2026-07-06 · thepragmaticquant_ The residual bootstrap assumes your model’s errors all come from the same distribution. Feed it a series whose noise grows over time and it quietly shuffles the loud errors in among the quiet ones, averaging the variance away, and hands back a standard error at three-quarters of the truth. Here is the failure on real numbers, and the wild bootstrap: a fix so simple it sounds like a joke. Keep… ## [Your bootstrap is lying to you](https://thepragmaticquant.com/your-bootstrap-is-lying-to-you/) _2026-07-06 · thepragmaticquant_ The ordinary bootstrap assumes your observations are independent. Feed it a time series and it quietly shuffles away the autocorrelation, then hands you a confidence interval too narrow to be true: a nominal 90% interval that covers the truth about half the time. Here is the failure on real numbers, and the block bootstrap that keeps the dependence and tells you most of the truth. ## [When is the swarm actually done?](https://thepragmaticquant.com/when-is-the-swarm-actually-done/) _2026-06-24 · thepragmaticquant_ Three services across a network, one dropped task, and a question nobody on the wire can answer: did the work actually get done? A copy-paste demo that exits 0 and prints SUCCESS while the real job dies unseen. ## [Refer a Friend, Earn Rewards (Sponsored)](https://crawlproof.com/a/O8nx5j48WQ3w) _2026-06-23 · **Sponsored**_ Earn rewards when a friend signs up for the Graphite Business Card ## [Task was destroyed but it is pending](https://thepragmaticquant.com/task-was-destroyed-but-pending/) _2026-06-24 · thepragmaticquant_ A teardown deadlock I diagnosed in Google's A2A reference SDK — Issue #1101, PR #1105 (417 added lines, 312 of them tests, CI green, merged upstream) — turned out to be one symptom of a primitive the whole agent ecosystem is missing: a sound signal that an async mesh has actually settled. ## [The first file an agent reads](https://thepragmaticquant.com/the-first-file-an-agent-reads/) _2026-06-19 · thepragmaticquant_ Coding agents read your library before they use it, and they start with the code — \`\_\_init\_\_.py\`, the type hints, the tool schemas — not your docs site. Here is how I made waitbus speak to that reader, and the one piece of documentation I deliberately did not ship. ## [Your AI coding agents can't hear each other — not even across vendors](https://thepragmaticquant.com/coding-agents-cant-hear-each-other/) _2026-06-14 · thepragmaticquant_ Your Claude Code, your Cursor, your tests, your CI — all on one box, none able to hear the others finish or fail. Here's the problem, and the proof: five real LLM agents on one bus, one fails, all wake. ## [How waitbus works: from event source to a waiting agent, over MCP](https://thepragmaticquant.com/source-to-subscriber-milliseconds/) _2026-06-14 · thepragmaticquant_ The architecture end to end — how an event gets from a source to a waiting agent in single-digit milliseconds, how an agent actually talks to the bus over MCP, and the decisions behind the build with what each one cost. ## [The numbers and the trust trail: benchmarking waitbus honestly](https://thepragmaticquant.com/why-my-benchmarks-lied/) _2026-06-14 · thepragmaticquant_ Two kinds of trust in one place — the benchmark methodology that makes the speed numbers survive a skeptic (Coordinated Omission, a bimodal p99, costs published as losses), and the supply-chain trail that lets you trust the artifact you install (SLSA provenance, sigstore, reproducible builds, and an honest list of the gaps). ## [Thirty-one papers, zero error analyses](https://thepragmaticquant.com/thirty-one-papers-zero-error-analyses/) _2026-06-12 · thepragmaticquant_ A sliding-window inner product updated one product at a time is the streaming similarity engine under motif and anomaly mining — its rounding error grows linearly in stream length, and across thirty-one matrix-profile papers nobody had ever done the forward-error math, or turned the fix into a config knob. ## [The accumulator that never moved](https://thepragmaticquant.com/the-accumulator-that-never-moved/) _2026-06-12 · thepragmaticquant_ An adversarial input where round-to-nearest throws away every increment in the same direction while the true sum climbs — and the coin flip that breaks the adversary's one weapon, measured across 20,000 seeds. ## [The error analysis everyone cites is for a kernel nobody runs](https://thepragmaticquant.com/the-kernel-nobody-analyzed/) _2026-06-12 · thepragmaticquant_ SCAMP and STUMPY accumulate a mean-centred covariance, not the textbook inner product — and the centred case is structurally different floating-point mathematics, not a special case at mu = 0. ## [Green is not evidence](https://thepragmaticquant.com/green-is-not-evidence/) _2026-06-12 · thepragmaticquant_ An experiment can validate the wrong quantity against the wrong envelope and pass, a self-test canary can fail identically whether the gate works or not, and an audit command can print nothing and exit 0 — three gate designs that read green while checking nothing. ## [Three axioms and one disclosed seam](https://thepragmaticquant.com/three-axioms-and-one-disclosed-seam/) _2026-06-12 · thepragmaticquant_ Four verification tools, one certificate: every machine-checked theorem mechanically reduced to three classical axioms with zero unfinished proofs — and the one place two provers cannot talk to each other, named right in the theorem statements. ## [Point a Domain, Start a Pit (Sponsored)](https://crawlproof.com/a/dAmYQF96swOF) _2026-06-11 · **Sponsored**_ Point any domain to a blacked-out, poison-green coming-soon page with an email waitlist. ## [One earthquake pinned my error bound forever](https://thepragmaticquant.com/one-earthquake-pinned-my-bound/) _2026-06-12 · thepragmaticquant_ A seismic stream, a global magnitude bound wrecked by a single mainshock, a windowed bound that heals — and a scoreboard where the headline empirical legs of two papers report zero violations.